ARTICLE

https://doi.org/10.1038/s41467-021-24818-x OPEN A distinct assembly pathway of the 39S late pre-mitoribosome ✉ ✉ Jingdong Cheng 1 , Otto Berninghausen 1 & Roland Beckmann 1

Assembly of the mitoribosome is largely enigmatic and involves numerous assembly factors. Little is known about their function and the architectural transitions of the pre-ribosomal intermediates. Here, we solve cryo-EM structures of the human 39S large subunit pre- fi 1234567890():,; , representing ve distinct late states. Besides the MALSU1 complex used as bait for affinity purification, we identify several assembly factors, including the DDX28 helicase, MRM3, GTPBP10 and the NSUN4-mTERF4 complex, all of which keep the 16S rRNA in immature conformations. The late transitions mainly involve rRNA domains IV and V, which form the central protuberance, the intersubunit side and the peptidyltransferase center of the 39S subunit. Unexpectedly, we find deacylated tRNA in the ribosomal E-site, suggesting a role in 39S assembly. Taken together, our study provides an architectural inventory of the distinct late assembly phase of the human 39S mitoribosome.

✉ 1 Center and Department for Biochemistry, LMU Munich, München, Germany. email: [email protected]; [email protected]

NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications 1 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x

he human mitochondrial 55S (mitoribosome) human mitochondria. Since mitoribosome assembly appears to Tconsists of a 28S small ribosomal subunit (mtSSU), formed be quite diverse between species, the pathway of human 39S by a 12S rRNA and 29 mitochondrial ribosomal mitoribosome assembly remains largely enigmatic. (MRPs), and the 39S large ribosomal subunit (mtLSU), formed by In this work, we use tagged assembly factors MALSU1 and a 16S rRNA and 50 MRPs. It is specialized to synthesize 13 GTPBP10 as baits for affinity purification to isolate late assembly hydrophobic polypeptides, which are essential for oxidative intermediates of the 39S LSU mitoribosome. Subsequent cryo-EM phosphorylation (OXPHOS), thus playing an essential role in single-particle analysis reveals the structures of nine late assembly maintaining the mitochondrial function. Numerous mutations in intermediates with distinct AF compositions and arrangements. mitochondrial rRNA, the ribosomal proteins or the translation We observe an overall conserved the rRNA maturation pathway regulators (initiation, elongation factors) are known to cause and several assembly factors apparently functioning to delay human diseases1,2. With the recent new development in cryo- folding of the rRNA. Moreover, unexpectedly a tRNA is present electron microscopy (cryo-EM), we have gained wealth infor- in the E site of two assembly intermediates. Together, our study mation about both the mtSSU and mtLSU structures, and also provides an architectural inventory of the distinct late assembly about the mode of translation by the active 55S ribosome3–12. phase of the human 39S mitoribosome. Comparing with its bacterial ancestor, 39S large ribosomal sub- units share most common structural units, such as six sub- domains in the 16S rRNA (domain I to VI), the peptide Results transferase center (PTC), L1 and L7/12 stalks, and a central Cryo-EM analysis of late mitochondrial ribosomal large sub- protuberance (CP). However, there are substantial differences in unit intermediates. Both, MALSU1 and GTPBP10, a member of structure and composition, such as an increased number of the Obg subfamily of GTPase and functionally conserved homolog mitochondrial ribosomal proteins and substantially different of E. coli ObgE, are expected to act during the late phases of rRNA content13, resulting in a high to RNA ratio in case mtLSU assembly23,24,30, and were thus chosen as baits for affinity of the human mitoribsome14. Moreover, a valine tRNA forms the purification. All particles were isolated using Flag-tag immuno- central protuberance, thereby replacing the 5S rRNA4. Taken precipitation from human HEK293 cells and characterized by together, this suggests a deviating assembly process for the cryo-EM single-particle analysis. After processing, we obtained mitochondrial ribosomes. nine well-defined classes that represent five different folding states Ribosome biogenesis in general involves the stepwise folding of the of the 16S rRNA with some additional variation regarding AF or ribosomal RNA and recruitment of ribosomal proteins with the help tRNA association (Fig. 1, Supplementary Figs. 1–3). Not surpris- of the assembly factors. Similar to the more extensively studied ingly, according to the local resolution estimation, they all dis- bacterial ribosome, the mitochondrial 39S large subunit requires played a highly ordered (high-resolution) solvent side, whereas a dozens of assembly factors (AFs), such as methyltransferases15–17, rather flexible intersubunit side was indicative of flexible regions pseudouridine synthases18,19,GTPases20–25,RNAhelicases26,27,and waiting for the late maturation events to happen (Supplementary other factors28–31. However, only a handful of homologs of these Fig. 4). However, all reconstructions displayed average resolutions factors have been captured on bacterial ribosome assembly inter- between 3.1 Å and 5.7 Å allowing us to build de novo models or to mediates in structural studies32. Unlike the bacterial ribosome, until perform rigid-body fitting of homology structures (Fig. 1, Sup- now there are only four highly-conserved modification sites mapped plementary Figs. 4 and 5, Supplementary Tables 1 and 2). on the mtLSU: one pseudouridylation at U1397, which is carried out According to the folding states of the 16S rRNA, we classified by RPUSD418,andthree2’-O-ribose methylations, which are cata- them into five principle states (named 1–5) and arranged them in lyzed by three methyltransferases, MRM1, MRM2, and MRM315. the most plausible order, most probably representing the MRM1 modifiesG1145duringtheveryearlymitoribosomeassembly sequence of the late maturation steps of the 39S ribosome. Two pathway, since it appears to modify naked RNA and does not co- major states contained sub-states due to the different AF sediment with pre-mitoribosomal particles15.Theothertwosites, composition (named A to D) (Fig. 1 and Supplementary Table 2). U1369 and G1370, are modified by MRM2 and MRM3, respectively, State 1 already exhibited a 39S ribosome-like structure showing and are located on the A-loop in which these modifications are both the L1 and L7/L12 stalk, however, with an immature central suggested to influence A site tRNA accommodation17. In addition, protuberance. In addition, rRNA helices H64–65 and H67–71 of NSUN4 has been identified as an rRNA m5C methyltransferase for domain IV, and helices H80–H93 of domain V of the 16S rRNA the methylation of 12S rRNA of the mtSSU on position C911. It were largely delocalized. State 2, which falls between State 1 and 3, belongs to the same protein family as the bacterial assembly factor displayed already close to mature positions of helices H80–88 of RsmB, however, in contrast to RsmB it cannot directly bind to RNA domain V of the 16S rRNA as well as rigid binding of bL33, without the help of mTERF4. mTERF4 belongs to the highly- leaving only helix H81 of the 16S rRNA unassigned (Fig. 1). From conserved family of mitochondrial transcription termination factors state 3 to 4, helices H80–H88 and H89–H93 of the domain V (mTERF), which usually have nucleic acid binding activity. Notably, were folded in place sequentially, together with the central the NSUN4–mTERF4 complex has also been involved in ribosomal protuberance. In addition, the NSUN4–mTERF4 complex was subunit joining in mitochondria16,29,33,34. observed interacting with the immature helices H68–71 of Recently, structures of two late assembly intermediates of the domain IV in state 4. Furthermore, state 4 showed bL36 in its human 39S mitoribosome were reported, in which one of them mature position. The folding state of the 16S rRNA in state 3B, has already adopted the mature 16S rRNA conformation31. Here, 3C, and 3D is the same as state 3A, yet, a major difference is the the MALSU1 complex (MALSU1, L0R8F8, and mt-ACP) was presence of a tRNA in the E-site (E-tRNA) and/or the MALSU1 identified in both intermediates. MALSU1 belongs to the RsfS complex. In detail, state 3A contained both the MALSU1 complex protein family which, similar to eIF6 in , binds to uL14 and the E-site tRNA, whereas state 3B and 3C contained either on the large subunit and serves as an anti-association factor the MALSU1 complex or tRNA (Fig. 1 and Supplementary preventing premature subunit joining in bacteria30,35. Moreover, Fig. 3). Surprisingly state 3D contained neither one (Supplemen- several large subunit assembly intermediates of kinetoplastid LSU tary Fig. 3). Similar to state 3, state 5 also divides into two sub- mitoribosome have also been solved by cryo-EM recently36–38. states, state 5A and 5B with E-tRNA only present in state 5A. Although, we share some of the common AFs with kinetoplastids, Notably, apart from the remaining association of E-tRNA and the the majority of their AFs are species specific and do not exist in MALSU1 complex, state 5A resembles already the mature 39S

2 NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x ARTICLE

State 1 State 2 State 3 State 4 State 5

3A 5A cp H81 L7/L12 stalk E-tRNA E-tRNA II Domain V Domain V V H80-88 H80, 82-88 L1 stalk IV VI DDX28 bL33 H89 III NSUN4 MRM3 3B 5B VI MRM3 GTPBP10 mTERF4 MALSU1 complex H90-92 IV

H89-93 H68-71

bL33, bL35 NSUN4, mTERF4, bL36 H80, 82-88 in place H81 in placeH34-35, H89-93 in place H67-71 in place

DDX28, MRM3 GTPBP10 NSUN4, mTERF4

IV V IV V

III III

II II VI VI I I

tRNAVal

Fig. 1 Ensemble of late human 39S mitoribosome assembly intermediates. Cryo-EM maps of seven different intermediates of the 39S mitoribosome (top row) are shown with the 16S rRNA secondary structures corresponding to states 1–5 (bottom row). All assembly factors (MRM3: blue and sandy brown, DDX28: green, GTPBP10: orange, NSUN4: red, mTERF4: green, MALSU1 complex: purple) and six subdomains of the 16S rRNA are color coded (domain 1–6: purple, blue, magenta, yellow, hot pink, and green) and ribosomal landmarks indicated (cp, central protuberance). Immature or unstructured regions of the rRNA are shaded in light blue and gray in the secondary structure schemes, respectively. All maps were filtered to local resolution using Relion. mitoribosome (Fig. 1). In conclusion, our structural analysis Although the local resolution is too limited as to reveal atomic detail, revealed several novel assembly intermediates representing the we can unambiguously rigid body fitahomologymodelofDDX28 late maturation steps of the human 39S mitoribosome. into our state 1 based on the resolved secondary structure (Supplementary Fig. 5a). The assignment of this density to DDX28 fi DDX28 keeps the central protuberance immature. Of the is in agreement with mass spec analysis con rming its presence in the fi analyzed intermediates, State 1 showed the least mature con- puri ed intermediates (Supplementary Data 1, 2). Notably, when formation, characterized by the immature central protuberance inspecting the substrate binding pocket of DDX28 we could clearly – and an immature PTC. However, at the same time it already trace an RNA density, corresponding to the region 1245 1251 of 16S showed the well-defined solvent side shell of the 39S mitoribo- rRNA domain V (Fig. 2b and Supplementary Fig. 5b). Thus, some (Fig. 1). Notably, a similar state has also been observed consistent with the published data, DDX28 is involved in late 39S both, in vivo and in vitro for bacterial 50S pre-ribosomes39–42.In mitoribosome assembly by exerting is helicase activity on the central 27 state 1, the core of the 39S mitoribosome, formed by domain I, II, protuberance . III, and VI of the 16S rRNA, was well resolved and properly Due to the binding of DDX28 to the central protuberance, – folded in its native mature conformation. Yet, besides rRNA helix helices H80 88 of domain V were kept in a substantially distant H75, which forms the L1 stalk, almost the entire domain IV and location from their mature position. Helix H86, when compared domain V were still delocalized (Fig. 1). with the mature conformation, appeared simply shifted upwards, The central protuberance consists mainly of helices H80–88 of whereas rRNA helices H81, H82, and H88 were also rotated domain V of the 16S rRNA and a valine tRNA. When comparing upwards, resulting in an open conformation of this entire state1withthelaterstate5A,inwhichthe16SrRNAalready subdomain (Fig. 2c). Two of the mitoribosomal proteins, bL33 adopted its mature conformation, the valine tRNA had already been and bL35, which interact tightly with the rRNA of this region in recruited, whereas the helices H80–88 of domain V are completely the mature conformation, were not yet recruited (Fig. 2d). distorted through interaction with the AF DDX28 (Fig. 2a). This AF Furthermore, the bound valine tRNA and the associating is a DEAD box helicase that has been shown to localize to the RNA mitoribosomal proteins were observed in a rotated premature granules where the mitoribosome assembly occurs. DDX28 is known conformation (Fig. 2e). Taken together, the presence of DDX28 in to interact with the 16S rRNA of the 39S mitoribosome and has been state 1 results in the stabilization of the central protuberance in an proposed previously to function during late assembly phases26,27. immature conformation.

NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications 3 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x

abc State 1 State 5A H82 H86 Val H82 H81 tRNA tRNAVal H82 H82 H81 RecA1 H88 H81 DDX28 H81 H82

H88 C1245 H86 RecA2 H81

State 1 A1251 DDX28 State 5A d State 1 State 1 State 5A e mL38 bL33 tRNAVal

H81 H86 H86 bL35 H82 H88

H82 H88 tRNAVal H88 H13 H13 bL27 State 1 State 5A

Fig. 2 The immature central protuberance and DDX28 helicase in state 1. a Overall positions of H80–88 of domain V of the 16S rRNA and DDX28. b DDX28 interacts with nucleotides 1245–1251 of domain V of the 16S rRNA. c Repositioning of H80–88 between state 1(pink) and 5A (gray). d Density showing void between H82 and H88 in state 1 (left), and models of state 1 and 5A showing incorporation of bL33 and bL35. e Comparison between state 5A and state 1, highlighting an immature conformation of the valine tRNA in the central protuberance of state 1.

MRM3 and GTPBP10 keep the PTC immature. Situated below of modification of the A-loop is conserved between species in the central protuberance and functionally most important in the mitochondria. In eukaryotes, like in mitochondria, it is executed large ribosomal subunit is the PTC region, which comprises the by a methyltransferase protein (Spb1 in yeast) during late rRNA helices H90-93 of domain V. In state 1, however, this assembly and not as many modifications by a snoRNP in an region is shifted outwards and located right in front of the earlier phase44. Apparently, in bacteria a relative of MRM3, FtsJ, immature helices H80–88 of domain V (Fig. 3a). This structural executes this modification, probably in a similar way45. arrangement is stabilized by the methyltransferase MRM3 dimer Also in state 1, helix H89 of domain V was still located in the (Fig. 3b). In addition, we observed a direct interaction between vicinity of its mature position, yet shifted outwards and stabilized the N-terminal domain of one copy of MRM3 and the RecA2 by GTPBP10 (Fig. 3d, e). Consistent with our structural analysis, domain of DDX28, which may indicate a coordination in the biochemical studies have shown previously that this AF directly recruitment of these two factors (Fig. 3c and Supplementary interacts with 16S rRNA between states containing DDX28 or Fig. 5c). MRM3 belongs to the SpoU methyltransferase family NSUN4/mTERF4, respectively23,24. We observed the GTPBP10 characterized by a classical C-terminal methyltransferase domain bound to the sarcin-ricin-loop (SRL) and the stalk base, which is as well as a structured N-terminal domain. Two SpoU-like in agreement with the findings for ObgE, the bacterial homolog of methyltransferases, MRM1 and MRM3, are present in human GTPBP10 (Fig. 3f)46. Notably, also in eukaryotes the ObgE-type cells and function in mitoribosome assembly, but only MRM3 has GTPase Nog1 binds to and displaces the rRNA helix H89 as been shown to modify G1370 in the A-loop in helix H92, which observed in yeast47. Thus, the GTPBP10/ObgE regulation may be agrees perfectly with their assigned position (Fig. 3b)15,17.In conserved and, like the translational GTPases, activation of GTP contrast, MRM1 is known to modify G1145, a residue that could hydrolysis of GTPBP10 might involve the SRL of the mtLSU. not be reached from the observed density15. Based on secondary However, ObgE, which was used as a bait to purify native E. coli structure, we can unambiguously fit MRM3 (Supplementary intermediates, was observed to associate only with later states of Fig. 5d), which positions the active center of methyltransferase assembly and thus has a different function in 50S subunit domains right on top of the A-loop (Fig. 3b). In agreement with assembly compared to GTPBP10 in 39S assembly42. this assignment, a homologous methyltransferase was observed also in the mitochondrial assembly intermediate from Trypano- Premature association of E-site tRNA in states 3 and 5. During soma brucei in an essentially identical conformation (Fig. 3b). the late maturation, we observed four different sub-states of the However, the location of rRNA domain V is dramatically dif- state 3, state 3A–3D. They all show the same immature con- ferent, which highlights the specificity of the human mitoribo- formation of the 16S rRNA as the previously reported 39S some assembly pathway (Supplementary Fig. 6)36. Although at mitoribosome intermediate31. However, only states 3A and 3B, the given resolution we can only tentatively assign a MRM3 but not 3C and 3D showed the MALSU1 complex. Surprisingly, homodimer to the observed symmetric density, it would be we also found premature incorporation of a tRNA into the E-site reminiscent to many members of the SpoU family methyl- in states 3A and 3C (Fig. 4). Due to steric hindrance, binding of transferases that form functional homodimers43 and is thus the the NSUN4–mTERF4 complex in state 4 prevents binding of the most plausible interpretation. In any case, we find that the mode E-site tRNA in this state (Fig. 1). Apparently, it can rebind in

4 NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x ARTICLE

a State 1 State 5A b mt-LAF6 MRM3 MRM3 H92 MRM3 H92

H92 H92 mt-LAF5 MRM3

H. s. T. b. cdef State 1 L7/L12 stalk

H89 H89 GTPBP10

H89 GTPBP10 SRL N terminus RecA2 H89 Obg GTPase MRM3 DDX28 domain domain GTPBP10

Fig. 3 State 1 displays an immature PTC with GTPBB10 and MRM3 complex bound. a Overall position of domain V in state 1 (left) and 5A (right). H92 of the 16S rRNA and MRM3 dimer (in blue and sandy brown) are labeled. b Comparison between the human MRM3 homodimer bound to H90–H92 and the mt-LAF5-mt-LAF6 heterodimer (in light blue and salmon, respectively) bound to H92 in maturing Trypanosoma brucei (PDB: 6YXX) large mitoribosomal subunit. c Interaction of N-terminal domain of MRM3 with the RecA2 domain of DDX28. d Overview of GTPBP10 (orange) and H89 (hot pink) in state 1. e H89 adopts an immature conformation in the presence of GTPBP10 (pink, H89 in state 1: gray, mature H89). f GTPBP10 interacting with H89 and the SRL of the 16S rRNA and the L7/L12 stalk. H. s.: Homo sapiens, T. b.: Trypanosoma brucei.

abc

ICT1 G1239 A1240

E-tRNA E-tRNA L1 stalk

A73 State 3A

Fig. 4 Binding of E-site tRNA to immature mitoribosomes. a State 3A shows an E-site tRNA (cyan) interacting with ICT1 (orange). b Cryo-EM density of the E-site tRNA superimposed on the model indicating flexibility of the anticodon stem. c Detailed view of the CCA end of the deacylated tRNA with the terminal A stacking between 16S rRNA bases. state 5A or, alternatively, a direct transition from state 3 to 5 is appeared already almost fully matured, leaving only helices possible. In any case, this finding contradicts the commonly H67–71 of the domain IV immature (Fig. 1). In addition, we accepted principle that, in general, premature binding of trans- observed an additional density in this state and, based on sec- lation factors or components is prevented during ribosome ondary structure information, we could unambiguously fitthe assembly. The observed binding of E-site tRNA could be acci- NSUN4–mTERF4 X-ray structure (Fig. 5aandSupplementary dental and without any functional relevance. Yet, another possi- Fig. 5e). Recent structures have shown that NSUN4 forms a bility is that mitochondria take advantage of the translational stable complex with mTERF4, suggesting that mTERF4 targets components already present in the mitochondrial matrix to either NSUN4 to the rRNA and regulates its activity33,34.Although stabilize assembly intermediates or to monitor the folding state of NSUN4 only modifies residue C911 in the mtSSU, it was also the central protuberance even before its full maturation. shown to interact with the mtLSU and, thus, has a dual function in both mtSSU and mtLSU assembly16. The NSUN4–mTERF4 complex grips the immature domain IV. Consistent with these published results, we observed the – In the observed state 4, the domain V of the 16S rRNA NSUN4 mTERF4 complex directly above the peptide transfer

NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications 5 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x

ab c NSUN4 NSUN4 H68-71 NSUN4 mTERF4

H82 SAM mTERF4 H82 17 Å H81 H80 H80 H75 H88 State 4 H86 de

H68-71 H68-71 NSUN4 NSUN4 NSUN4

H68-71 mTERF4

mTERF4 mTERF4 -70 7

Fig. 5 Immature rRNA helices H68–71 and NSUN4–mTERF4 complex in state 4. a View onto the intersubunit side of the map (top) and model (bottom) of state 4. The NSUN4–mTERF4 complex is shown as a surface model. b, c close-up views of the interaction between the NSUN4–mTERF4 complex and the 16S rRNA, illustrating the distance between the catalytic center of NSUN4 and the 16S rRNA. To locate the active center of NSUN4 S-adenosyl methionine (SAM) was docked from the crystal structure (pdb:4FZV). d rRNA helices H68–71 interact with mTERF4 which is shown as surface and colored according to its surface potential. e Comparison of unbound mTERF4 (PDB: 4FZV, gray) with mTERF4 bound to the mitoribosome indicating conformational changes. Helices H68–71 in (a, d) are shown as yellow density with dashed lines. center (PTC), with the NSUN4 subunit very close to the helix Discussion H80–88 region of domain V and the mTERF4 subunit Based on the conformation and maturation states of the 16S sandwiched between helices H68–71 and H75 of domain IV rRNA, our ensemble of intermediate structures purified with the (Fig. 5a, b). The active site pocket of NSUN4 was located in assembly factors MALSU1 and GTPBP10 can be put in a juxtaposition to helix H82 of the 16S rRNA, however, with a sequential order representing the late maturation path of the too large distance as to reach it for modification (Fig. 5c). This human 39S mitoribosome (Fig. 6). In the early intermediate state is in agreement with the finding that NSUN4 methylate 12S 1, the terminal domains of 16S rRNA, domains I, II, III, and VI but not 16S rRNA16.Notably,inthisconformationNSUN4 first assemble to form the solvent side core of the 39S mitor- would also be prevented from accessing the 12S rRNA of the ibosome, only leaving the middle domains IV and V immature. 28S mitoribosome, thereby excluding modification activity as This is achieved with the employment of assembly factors a potential assembly check-point upon subunit joining. Thus, DDX28, MRM3, and GTPBP10, which each bind and stabilize the NSUN4–mTERF4 complex may rather have a scaffolding different parts of these two rRNA domains in immature con- function on the mtLSU and likely serves to check and stabilize formations, respectively. Upon further progress, DDX28 together the maturation state of the domain V of the 16S rRNA while with the MRM3 complex leaves the intermediate with the helices keeping domain IV in a largely immature position. Specifi- H80–88 of domain V matured. Thus, the further matured con- cally, the mTERF4 subunit maintains helices H68–71 of formation of the central protuberance may be checked and sta- domain IV in an immature conformation. This RNA segment bilized by binding of the E site tRNA. In this state (state 3), is the last building block required to complete the PTC. As although we could not rule out that GTPBP10 might be dis- suggested before, it is likely that a positively charged patch on sociated during preparation, GTPBP10 is expected to still be the surface of mTERF4 is used to interact with this region associated with the flexible and therefore invisible 16S rRNA helix (Fig. 5d)34. When comparing mTERF4 in our complex with H89 since the same state was obtained using different approaches. the RNA-free X-ray structure, we observed that the N- Thus, GTPBP10 most likely moves together with H89 without terminus of mTERF4 undergoes a conformational change providing visible density, but it is still associated with the parti- towards a more opened conformation (Fig. 5e).Thereby,the cles. Subsequently, upon dissociation of GTPBP10, which keeps domain IV rRNA is held in the concave middle of mTERF4, helices H89–93 of domain V immature, allows for the maturation thus keeping this region far away from its mature location. of domain V to conclude, leaving only helices H68–71 of domain

6 NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x ARTICLE

H81

V NSUN4/mTERF4 DDX28 Domain V bL33, NSUN4 H80,82-88 I II bL35 bL36 MRM3 mTERF4 IV

GTPBP10 MRM3 Domain V H89-93 III VI GTPBP10 DDX28 MRM3 MRM3

State 1 State 2 State 3State 4 State 5

Fig. 6 Late assembly steps of the 39S mitoribosome. Cartoon depicting the late assembly transitions during the 39S mitoribosome maturation. A possible bypass from state 3 directly to state 5 is indicated with dash line. Assembly factors and the six subdomains (I–VI) of the 16S rRNA are color coded. For detailed description see text.

IV in an immature conformation. This intermediate state is Thus, we can fit either homology models or crystal structures of stabilized by the interaction with the NSUN4–mTERF4 complex, the respective factors. NSUN4 and mTERF4, for example, form a which at the same time, can also serve as a final check-point to heterodimer and here the X-ray structure of the entire complex monitor the folding of the domain V. At this point, only dis- fits the density after small adjustments (Supplementary Fig. 5). sociation of the NSUN4–mTERF4 complex is required to allow Third, the DDX28 density clearly adopts a DEAD box helicase for the final rearrangement of rRNA domain IV and thus for- fold and DDX28 is the only DEAD box helicase known to be mation of the mature 39S mitoribosome. The observed ensemble involved and present in the MS analysis. Fourth, it is known that of intermediates provides little evidence for alternative pathways MRM3 modifies G1370 on the A loop which is in perfect when focusing on the conformational transitions of the rRNA, agreement with the position of the MRM3 dimer density after however, we cannot exclude that they exist, possibly under stress docking15,17. Last, the GTPBP10 density has an ObgE type conditions. Only the transition from state 3 to 5 may happen GTPase shape and interacts with the L7/L12 stalk and SRL, which directly by bypassing state 4, which would imply that maturation is commonly observed for this protein family both, in of remaining immature rRNA domains IV and V can occur and eukaryotes. MTG2 as a second mitochondrial ObgE-like concomitantly with the MTERF4–NSUN4 complex being GTPase also participates in 39S mitoribosome assembly, however, dispensable. it is not present in our sample according to MS analysis (Sup- This ends-to-middle order of rRNA domain accommodation is plementary Data 1, 2). We are thus confident that our density conserved in 23S/25S rRNA assembly of the cytoplasmic 50S/60S assignments are highly likely to be correct. subunit maturation, in which domains IV and V are also kept Regarding the biogenesis of the eukaryotic 80S ribosome, it is a immature until the late stages, with H69 of the 25S rRNA being well-accepted concept that during the assembly all functional the last to be completed32,48,49. A similar non-canonical sequence regions are kept immature or masked in order to prevent pre- has also been observed with 18S rRNA assembly of the cyto- mature association of translation factors, such as mRNA or plasmic 40S small subunit maturation50. Taking this into account, tRNA. To our surprise, however, this appears not to be valid for we conclude that an initial assembly sequence of ribosomal RNA the mitoribosome assembly since we observed the premature assembly, which is seeded by the 5′ and 3′ ends, presents a uni- incorporation of mature deacylated tRNA in the E site in two of versal mechanism, conserved from prokaryotes to eukaryotes and the intermediates, which only been observed on mature 39S within eukaryotes shared between cytoplasmic and organellar mitoribosome so far51. One possibility is that such early binding (mitochondrial) ribosomes (Supplementary Table 3). We spec- of tRNA serves a test drive for functionality of the E-site, similar ulate that one reason for this to occur is the implementation of an to the suggested test drive of the pre-40S by engaging the mature efficient and reliable way to ensure the completeness of the rRNA 60S subunit to from a first complete 80S-like ribosome in yeast52. transcription, thereby mitigating unnecessary energy expenditure However, when taking also the replacement of 5S rRNA by the on incomplete rRNA intermediates. The same principle may also valine tRNA in the central protuberance into account, it appears apply to the 12S rRNA assembly of the 28S small mitoribosomal that mitoribosome architecture and assembly has evolved to subunit. maximize the usage of factors that are already present in mito- Through structural analysis, we identified DDX28, MRM3, chondria. This limits the requirements for encoded in the GTPBP10, NSUN4, and mTERF4 as late assembly factors on the mitochondrial DNA (e.g., for the 5S rRNA) and also for the 39S mitoribosomal intermediates. Although we do not have suf- import of nuclear-encoded factors. In agreement with this idea, ficient resolution to confirm all the assignments based on side we also observe the repurposing of the NSUN4–mTERF4 chain information, we are confident in our assignments for the methyltransferase complex. This enzyme is primarily active in the following reasons: First, the presence of all factors in the isolated modification of the small subunit rRNA16. But instead of dis- intermediates was confirmed by mass spectrometry (MS) (Sup- playing the function as a methyltransferase as for the 12S rRNA, plementary Data 1, 2). Second, our maps with density corre- on the 16S rRNA it may only sense the maturation state of the sponding to DDX28, MRM3, and the NSUN4–mTERF4 regions large subunit and triggers the final maturation step of domain IV have sufficient local resolution to assign the secondary structure. rRNA. Interestingly, we observed before that RNA modifying

NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications 7 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x enzymes can be repurposed in order to act on another ribosome original frames were aligned, summed and drift-corrected using MotionCor254. biogenesis substrate without displaying its original enzymatic Contrast transfer function parameters and resolution were estimated for each 55 56 activity. Nop1, for example, exerts methyltransferase activity, yet, micrograph using CTFFIND4 and Gctf , respectively. Micrographs with an estimated resolution below 4 Å and astigmatism below 5% were manually screened in the yeast 90S pre-ribosome, Nop1 also cannot access to the 18S for contamination or carbon rupture. As a result, a total 11,388 micrographs (for rRNA, thus can only serve as a scaffold protein to maintain the MALSU1 sample) and 9452 micrographs (for GTPBP10 sample) were selected. U3 snoRNP structure for coordinating the pre-40S ribosome53.In Particle picking was carried out using Gautomatch with a low pass filtered mature 4 fi conclusion, a reoccurring theme during ribosome biogenesis is human 39S mitoribosome as a reference . After 2D classi cation, a total 785,354 particles (for MALSU1 sample) and 1,813,074 particles (for GTPBBP10 sample) the evolution of modifying enzymes to adopt additional roles were submitted to 3D refinement, 3D classification, CTF refinement as shown in beyond their enzymatic activity, mostly gaining functions as Supplementary Fig. 1 using Relion 3.157. During the revision of this paper, we rRNA chaperones or scaffold proteins. collected another dataset using uL17 as bait. This dataset will be discussed some- Recently, several kinetoplastid mitoribosomal assembly inter- where else. However, a small portion of the dataset also contains the desired state 1. Together with data of state 1 from the other two datasets, we combined all of them mediates were reported, which allow for the comparison of together, and did another round of 3D classification focusing on MRM3/DDX28 mitoribosome assembly pathways between species (Supplemen- region. The resulting class which shown the best resolved MRM3/DDX28 region – tary Table 4)36 38. Despite the presence of some highly-conserved was picked to get the final reconstruction of state 1 (Supplementary Fig. 2). All the factors, the kinetoplastid uses numerous species-specific assembly final map was filtered according to the local resolution in Relion. factors which are the main contributors to shape the pre- ribosomal intermediates. As a result, the immature arrangement Model building and refinement. In general, the models of human mature 39S mitoribosome (PDB:3J9M)4 and two intermediates (PDB:5OOL, 5OOM)31 were used of 16S rRNA in our state 1 is completely different compared to for initial rigid body docking into the density. Since the 39S mitoribosome part of all the kinetoplastid intermediates, indicating substantial differences states resembled very well the mature 39S mitoribosome, the mature human 39S in the mitoribosome assembly pathways between species (Sup- mitoribosome model could be fitted as rigid body. The immature region in rRNA plementary Fig. 6). This is in line with the observation that due to domain IV and V were either simply removed or manually adjusted in Coot58 to better fi fi distinct evolutionary factors and in contrast to the cytoplasmic t the density. The MALSU1 complex was taken from PDB:5OOL and rigid body tted into every state31. For state 1, the overall resolution was not sufficient to build a 80S ribosome, mitochondrial ribosomes display dramatic differ- molecular model. DDX28 was fitted as a rigid body using a homology model generated ences between species. Yeast and plant mitoribosomes can even using the SWISS-MODEL server based on PDB: 4W7S59,60.Afterfitting, small be larger than cytoplasmic 80S ribosomes, whereas some insect adjustments were done manually in Coot. MRM3 and GTPBP10 were also rigid body and vertebrate mitoribosomes, such as the human one, are docked as homology model generated using the SWISS-MODEL server based on PDB: 4X3M and 1LNZ61, respectively. However, due to the low local resolution, no adjust- extremely small and retained only their rRNA core. Our study ments could be further done, and, in the final model of state 1, only poly-Ala models are thus provides information on the provided.Inthestates3Aand5A,theE-sitetRNAdidnothavesufficient resolution to assembly pathway which is specific for human cells and can be build a molecular model, thus a previously observed mitochondrial tRNA was used to directly related to the context of human mitochondrial disease. rigid body fit into the density9. The position A73 was manually built in Coot. For state 4, the crystal structure of NSUN4–mTERF4 complex was used to dock into the map, with a small manual adjustment of mTERF433,34.Alsohere,duetolowlocalresolution Methods of these two factors, only poly-alanine models were provided. Model refinement was Molecular cloning and generation of cell lines. Human MALSU1 and GTPBP10 carried out using Phenix.real_space_refine, and final models were validated using genes were amplified from a human cDNA library and inserted into a modified MolProbity62,63 inside of the Phenix software suit. Maps and models were visualized pcDNA5/FRT/TO vector which has a C-terminal Strep-Flag tag. The commercial and figures created with ChimeraX64. HEK Flp-In 293 T-Rex cell line was used to generate stable cell lines. In detail, the cells were split one day before the transfection, and transfected with the MALSU1 Reporting summary. Further information on research design is available in the Nature or GTPBP10 plasmid and a helper plasmid pOG44 using PEI according to the Research Reporting Summary linked to this article. manufacturer’s guidance. Two days after transfection, cells were split and trans- fered into a new plate with normal DMEM medium supplied with 10% FBS, 200 μg/ml hygromycin B, 10 μg/ml blasticidin and 1× penicillin/streptomycin. After Data availability several days of selection, the remaining cells were tested for protein expression and The EM density maps have been deposited in the Electron Microscopy Data Bank under frozen in liquid nitrogen. accession codes EMD-12919 (state 1), EMD-12920 (state 2), EMD-12921 (state 3A), EMD-12922 (state 3B), EMD-12923 (state 3C), EMD-12924 (state 3D), EMD-12925 Sample purification. The MALSU1 or GTPBP10 cells were cultured in normal (state 4), EMD-12926 (state 5A), EMD-12927 (state 5B), and the coordinates of the EM- DMEM medium with 10% FBS. To induce the expression of the tagged protein, based models have been deposited in the Protein Data Bank under accession codes PDB one day after splitting, a final concentration 1 μg/ml of tetracycline was added to 7OI6 (state 1), PDB 7OI7 (state 2), PDB 7OI8 (state 3A), PDB 7OI9 (state 3B), PDB induce expression for 24 h before harvest. A total of 25 dishes (15 cm) of cells were 7OIA (state 3C), PDB 7OIB (state 3D), PDB 7OIC (state 4), PDB 7OID (state 5A), PDB scratched down and washed once with cold PBS buffer. Lysis buffer (50 mM 7OIE (state 5B). Other coordinates of the used models in this study are available under HEPES pH 7.4, 150 mM KCl, 5 mM MgCl2, 1 mM DTT, 0.5 mM NaF, 1 mM accession codes: PDB 3J9M,PDB5OOL,PDB5OOM, PDB 4W7S, PDB 4X3M, and PDB Na3V3O4, 1× protease inhibitor mix) was added to resuspend the cell. After lysis 1LNZ. A Life Sciences Reporting Summary for this paper is available. Supplementary with 10 strokes using a douncer, the cell lysate was kept on ice for 10 min. Sub- datasets 1 and 2 are provided with this paper. sequently, the cell lysate was cleared by centrifugation at 10,000 × g at 4 °C for 25 min and immediately incubated with pre-equilibrated anti-Flag affinity beads for 2 h. After incubation, the beads were transferred into a new small column and first Received: 17 February 2021; Accepted: 1 July 2021; washed once with one column volume lysis buffer and then three times with wash buffer (20 mM HEPES pH 7.4, 150 mM KOAc, 5 mM Mg(OAc)2, 1 mM DTT). After washing, elution was performed by three times incubation with 200 μl elution buffer (0.2 mg/ml 3×Flag peptide in washing buffer) for 15 min. The final elution was concentrated using 100 kDa cut-off Amicon concentrator, and the final con- centration was measured using a NanoDrop photometer. References 1. De Silva, D., Tu, Y. T., Amunts, A., Fontanesi, F. & Barrientos, A. Electron microscopy and image processing. Before freezing, a final concentra- Mitochondrial ribosome assembly in health and disease. Cell Cycle 14, – tion of 0.05% Nikkol was added to improve the ice quality. 3.5 μl of the sample 2226 2250 (2015). = = ’ (MALSU1 sample: OD260 0.8, GTPBP10 sample: OD260 1.5) was directly 2. Ferrari, A., Del Olio, S. & Barrientos, A. The diseased mitoribosome. FEBS applied onto 2 nm carbon pre-coated R3/3 holey copper grids (Quantifoil), blotted Lett. https://doi.org/10.1002/1873-3468.14024 (2020). for 3 s and plunge-frozen in liquid ethane using a Vitrobot Mark IV. Cryo-EM data 3. Kaushal, P. S. et al. Cryo-EM structure of the small subunit of the mammalian was acquired on a Titan Krios transmission electron microscope (Thermo Fisher mitochondrial ribosome. Proc. Natl Acad. Sci. USA 111, 7284–7289 (2014). Scientific) operated at 300 kV under low-dose conditions (28 e− Å−2 in total) with 4. Amunts, A., Brown, A., Toots, J., Scheres, S. H. W. & Ramakrishnan, V. The a nominal pixel size of 1.084 Å on the object scale using EPU 2 software. A total structure of the human mitochondrial ribosome. Science 348,95–98 (2015). 12,430 micrographs (for MALSU1 sample) and 9834 micrographs (for 5. Greber, B. J. et al. The complete structure of the 55S mammalian GTPBP10 sample) were collected on a Falcon II direct electron detector. The mitochondrial ribosome. Science 348, 303–308 (2015).

8 NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x ARTICLE

6. Koripella, R. K. et al. Structure of human mitochondrial translation initiation 35. Wanschers, B. F. et al. C7orf30 specifically associates with the large subunit of factor 3 bound to the small ribosomal subunit. iScience 12,76–86 (2019). the mitochondrial ribosome and is involved in translation. Nucleic Acids Res. 7. Koripella, R. K., Sharma, M. R., Risteff, P., Keshavan, P. & Agrawal, R. K. 40, 4040–4051 (2012). Structural insights into unique features of the human mitochondrial ribosome 36. Jaskolowski, M. et al. Structural insights into the mechanism of mitoribosomal recycling. Proc. Natl Acad. Sci. USA 116, 8283–8288 (2019). large subunit biogenesis. Mol. cell 79, 629–644 e624 (2020). 8. Aibara, S., Singh, V., Modelska, A. & Amunts, A. Structural basis of 37. Soufari, H. et al. Structure of the mature kinetoplastids mitoribosome and mitochondrial translation. eLife https://doi.org/10.7554/eLife.58362 (2020). insights into its large subunit biogenesis. Proc. Natl Acad. Sci. USA 117, 9. Desai, N. et al. Elongational stalling activates mitoribosome-associated quality 29851–29861 (2020). control. Science 370, 1105–1110 (2020). 38. Tobiasson, V. et al. Interconnected assembly factors regulate the biogenesis of 10. Koripella, R. K. et al. Structures of the human mitochondrial ribosome bound mitoribosomal large subunit. EMBO J. 40, e106292 (2021). to EF-G1 reveal distinct features of mitochondrial translation elongation. Nat. 39. Davis, J. H. et al. Modular assembly of the bacterial large ribosomal subunit. Commun. 11, 3830 (2020). Cell 167, 1610–1622 e1615 (2016). 11. Kummer, E. & Ban, N. Structural insights into mammalian mitochondrial 40. Li, N. et al. Cryo-EM structures of the late-stage assembly intermediates of the translation elongation catalyzed by mtEFG1. EMBO J. 39, e104820 (2020). bacterial 50S ribosomal subunit. Nucleic Acids Res. 41, 7073–7083 12. Khawaja, A. et al. Distinct pre-initiation steps in human mitochondrial (2013). translation. Nat. Commun. 11, 2932 (2020). 41. Nikolay, R. et al. Structural visualization of the formation and activation of the 13. van der Sluis, E. O. et al. Parallel structural evolution of mitochondrial 50S ribosomal subunit during in vitro reconstitution. Mol. Cell 70, 881–893 ribosomes and OXPHOS complexes. Genome Biol. Evol. 7, 1235–1251 (2015). e883 (2018). 14. Sharma, M. R. et al. Structure of the mammalian mitochondrial ribosome 42. Nikolay, R. et al. Snapshots of native pre-50S ribosomes reveal a biogenesis reveals an expanded functional role for its component proteins. Cell 115, factor network and evolutionary specialization. Mol. Cell 81, 1200–1215 e1209 97–108 (2003). (2021). 15. Lee, K. W. & Bogenhagen, D. F. Assignment of 2’-O-methyltransferases to 43. Krishnamohan, A. & Jackman, J. E. A family divided: distinct structural and modification sites on the mammalian mitochondrial large subunit 16S mechanistic features of the SpoU-TrmD (SPOUT) methyltransferase ribosomal RNA (rRNA). J. Biol. Chem. 289, 24936–24942 (2014). superfamily. Biochemistry 58, 336–345 (2019). 16. Metodiev, M. D. et al. NSUN4 is a dual function mitochondrial protein 44. Lapeyre, B. & Purushothaman, S. K. Spb1p-directed formation of Gm2922 in required for both methylation of 12S rRNA and coordination of the ribosome catalytic center occurs at a late processing stage. Mol. cell 16, mitoribosomal assembly. PLoS Genet. 10, e1004110 (2014). 663–669 (2004). 17. Rorbach, J. et al. MRM2 and MRM3 are involved in biogenesis of the large 45. Caldas, T. et al. The FtsJ/RrmJ heat shock protein of Escherichia coli is a 23 S subunit of the mitochondrial ribosome. Mol. Biol. cell 25, 2542–2555 (2014). ribosomal RNA methyltransferase. J. Biol. Chem. 275, 16414–16419 18. Antonicka, H. et al. A pseudouridine synthase module is essential for (2000). mitochondrial protein synthesis and cell viability. EMBO Rep. 18,28–38 46. Feng, B. et al. Structural and functional insights into the mode of action of a (2017). universally conserved Obg GTPase. PLoS Biol. 12, e1001866 (2014). 19. Zaganelli, S. et al. The pseudouridine synthase RPUSD4 is an essential 47. Wu, S. et al. Diverse roles of assembly factors revealed by structures of late component of mitochondrial RNA granules. J. Biol. Chem. 292, 4519–4532 nuclear pre-60S ribosomes. Nature 534, 133–137 (2016). (2017). 48. Kater, L. et al. Visualizing the assembly pathway of nucleolar Pre-60S 20. Kotani, T., Akabane, S., Takeyasu, K., Ueda, T. & Takeuchi, N. Human G- ribosomes. Cell 171, 1599–1610 e1514 (2017). proteins, ObgH1 and Mtg1, associate with the large mitochondrial ribosome 49. Gamalinda, M. et al. A hierarchical model for assembly of eukaryotic 60S subunit and are involved in translation and assembly of respiratory complexes. ribosomal subunit domains. Genes Dev. 28, 198–210 (2014). Nucleic Acids Res. 41, 3713–3722 (2013). 50. Cheng, J. et al. Thermophile 90S pre-ribosome structures reveal the reverse 21. Hirano, Y., Ohniwa, R. L., Wada, C., Yoshimura, S. H. & Takeyasu, K. Human order of co-transcriptional 18S rRNA subdomain integration. Mol. cell 75, small G proteins, ObgH1, and ObgH2, participate in the maintenance of 1256–1269 e1257 (2019). mitochondria and nucleolar architectures. Genes Cells 11, 1295–1304 (2006). 51. Amunts, A. et al. Structure of the yeast mitochondrial large ribosomal subunit. 22. Maiti, P., Antonicka, H., Gingras, A. C., Shoubridge, E. A. & Barrientos, A. Science 343, 1485–1489 (2014). Human GTPBP5 (MTG2) fuels mitoribosome large subunit maturation by 52. Strunk, B. S., Novak, M. N., Young, C. L. & Karbstein, K. A translation-like facilitating 16S rRNA methylation. Nucleic Acids Res. 48, 7924–7943 (2020). cycle is a quality control checkpoint for maturing 40S ribosome subunits. Cell 23. Maiti, P., Kim, H. J., Tu, Y. T. & Barrientos, A. Human GTPBP10 is required 150, 111–121 (2012). for mitoribosome maturation. Nucleic Acids Res. 46, 11423–11437 (2018). 53. Cheng, J., Kellner, N., Berninghausen, O., Hurt, E. & Beckmann, R. 3.2-A- 24. Lavdovskaia, E. et al. The human Obg protein GTPBP10 is involved in resolution structure of the 90S preribosome before A1 pre-rRNA cleavage. mitoribosomal biogenesis. Nucleic Acids Res. 46, 8471–8482 (2018). Nat. Struct. Mol. Biol. 24, 954–964 (2017). 25. Cipullo, M. et al. Human GTPBP5 is involved in the late stage of 54. Zheng, S. Q. et al. MotionCor2: anisotropic correction of beam-induced mitoribosome large subunit assembly. Nucleic Acids Res. 49, 354–370 (2021). motion for improved cryo-electron microscopy. Nat. Methods 14, 331–332 26. Antonicka, H. & Shoubridge, E. A. Mitochondrial RNA granules are centers (2017). for posttranscriptional RNA processing and ribosome biogenesis. Cell Rep. 10, 55. Rohou, A. & Grigorieff, N. CTFFIND4: Fast and accurate defocus estimation 920–932 (2015). from electron micrographs. J. Struct. Biol. 192, 216–221 (2015). 27. Tu, Y. T. & Barrientos, A. The human mitochondrial DEAD-box protein 56. Zhang, K. Gctf: Real-time CTF determination and correction. J. Struct. Biol. DDX28 resides in RNA granules and functions in mitoribosome assembly. 193,1–12 (2016). Cell Rep. 10, 854–864 (2015). 57. Zivanov, J. et al. New tools for automated high-resolution cryo-EM structure 28. Wredenberg, A. et al. MTERF3 regulates mitochondrial ribosome biogenesis determination in RELION-3. eLife https://doi.org/10.7554/eLife.42166 in invertebrates and mammals. PLoS Genet. 9, e1003178 (2013). (2018). 29. Camara, Y. et al. MTERF4 regulates translation by targeting the 58. Emsley, P. & Cowtan, K. Coot: model-building tools for molecular graphics. methyltransferase NSUN4 to the mammalian mitochondrial ribosome. Cell Acta Crystallogr. Sect. D. Biol. Crystallogr. 60, 2126–2132 (2004). Metab. 13, 527–539 (2011). 59. Biasini, M. et al. SWISS-MODEL: modelling protein tertiary and quaternary 30. Fung, S., Nishimura, T., Sasarman, F. & Shoubridge, E. A. The conserved structure using evolutionary information. Nucleic Acids Res. 42, W252–W258 interaction of C7orf30 with MRPL14 promotes biogenesis of the (2014). mitochondrial large ribosomal subunit and mitochondrial translation. Mol. 60. Jacewicz, A., Schwer, B., Smith, P. & Shuman, S. Crystal structure, mutational Biol. Cell 24, 184–193 (2013). analysis and RNA-dependent ATPase activity of the yeast DEAD-box pre- 31. Brown, A. et al. Structures of the human mitochondrial ribosome in native mRNA splicing factor Prp28. Nucleic Acids Res. 42, 12885–12898 (2014). states of assembly. Nat. Struct. Mol. Biol. 24, 866–869 (2017). 61. Buglino, J., Shen, V., Hakimian, P. & Lima, C. D. Structural and biochemical 32. Davis, J. H. & Williamson, J. R. Structure and dynamics of bacterial ribosome analysis of the Obg GTP binding protein. Structure 10, 1581–1592 (2002). biogenesis. Philos. Trans. R Soc. Lond. B Biol. Sci. 372, 20160181 (2017). 62. Adams, P. D. et al. PHENIX: a comprehensive Python-based system for 33. Spahr, H., Habermann, B., Gustafsson, C. M., Larsson, N. G. & Hallberg, B. M. macromolecular structure solution. Acta Crystallogr. Sect. D., Biol. Crystallogr. Structure of the human MTERF4-NSUN4 protein complex that regulates 66, 213–221 (2010). mitochondrial ribosome biogenesis. Proc. Natl Acad. Sci. USA 109, 63. Chen, V. B. et al. MolProbity: all-atom structure validation for 15253–15258 (2012). macromolecular crystallography. Acta Crystallogr. Sect. D., Biol. Crystallogr. 34. Yakubovskaya, E. et al. Structure of the essential MTERF4:NSUN4 protein 66,12–21 (2010). complex reveals how an MTERF protein collaborates to facilitate rRNA 64. Goddard, T. D. et al. UCSF ChimeraX: meeting modern challenges in modification. Structure 20, 1940–1947 (2012). visualization and analysis. Protein Sci. 27,14–25 (2018).

NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications 9 ARTICLE NATURE COMMUNICATIONS | https://doi.org/10.1038/s41467-021-24818-x

Acknowledgements Peer review information Nature Communication thanks Jose Llacer, John Woolford, The authors thank S. Rieder, C. Ungewickell, and A. Gilmozzi for technical assistance, and other, anonymous, reviewers for their contributions to the peer review of this work. L. Kater for discussions and critical comments on the manuscript. This research was Peer review reports are available. supported by grants from the Deutsche Forschungsgemeinschaft (GRK1721) and by an European Research Council (ERC) Advanced Grant (HumanRibogenesis) to R.B. Reprints and permission information is available at http://www.nature.com/reprints

’ Author contributions Publisher s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. J.C. and R.B. designed the study. J.C. performed all the experiments and O.B. collected cryo-EM data. J.C. and R.B. analyzed the structures, interpreted results, and wrote the manuscript. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, Funding adaptation, distribution and reproduction in any medium or format, as long as you give Open Access funding enabled and organized by Projekt DEAL. appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party Competing interests material in this article are included in the article’s Creative Commons license, unless The authors declare no competing interests. indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from Additional information the copyright holder. To view a copy of this license, visit http://creativecommons.org/ Supplementary information The online version contains supplementary material licenses/by/4.0/. available at https://doi.org/10.1038/s41467-021-24818-x.

Correspondence and requests for materials should be addressed to J.C. or R.B. © The Author(s) 2021

10 NATURE COMMUNICATIONS | (2021) 12:4544 | https://doi.org/10.1038/s41467-021-24818-x | www.nature.com/naturecommunications